Practical Bot Development

Szymon Rozga has worked in the software development space for over 14 years taking a shot at everything from Rails to C to Java to Python to .Net. He found a passion working on Windows Desktop trading applications on Wall Street with a focus on the front end user experience. The interest in attention to user interface details would take him on a tour of the different UI technologies on the Windows, Web and iOS/Android platforms. He enjoys leading deep discovery sessions and scoping solutions for his clients. Szymon has managed teams of engineers on a variety of projects and can be commonly found leading engagements, mentoring junior engineers (another passion of his) and collaborating on building team engineering capacity in emerging technologies. Starting in 2016, Szymon has spent all his time dedicated to building bots for clients whether the bot runs on a mobile app, the web or a messaging platform. He has built a practice around the technology and is involved in a handful of bot projects at any time. He has a passion for paying attention to conversational experience details and is a big proponent of getting your bot to do a few tasks really well. He has a deep interest in all things Machine Learning and Artificial Intelligence. In his spare time he enjoys walking in nature, relaxing on the beach, playing guitar and spending time with his wife Kimberly, his son Theodore and his dog Chelsea. He's an avid reader, especially of things completely unrelated to software development.

Chapter 1: Introducing Bots

Motivation: apps vs bots and interacting with computers using Natural Language Interfaces. Google Now, Siri, Alexa. Great progress in recent years in both translation, voice recognition and natural language understanding technologies. Democratization of AI and Machine Learning APIs. Bots as a new way of implementing user interfaces.

Use Case: describe a bot to help users order tickets

Use it to describe the following concepts:



Natural language recognition
Text conversations
Rich media conversations
Bot-led conversations
Human hand off
Embedding into messaging apps

Chapter 2: Bot User Design (UX)


Motivation: there is a good way for bots to communicate to you and there is a bad way. This chapter will distill some common "good" approaches to a bot and Also show some examples of bad communication.

conversational User Experience




Text vs rich conversations

Bot initiated conversations

Perceived performance

Personalized content

Concise and clear messaging

Consistent voice

Do one thing well, provide value


Chapter 3: Natural Language Understanding


Motivation: part of the reason bots are becoming a thing is that natural language recognition technologies are widely available and easier to use.

Intent classification

Entity Extractors - Named Entity Extraction

LUIS




How to create a model

How to train LUIS to recognize intents and entities (use ticket ordering bot as a use case)




Action Fulfillment - not used in LUIS anymore, but still used in cloud based, intent-led bot frameworks. It is interesting to know when creating bots.


Chapter 4: Introducing Microsoft Bot Builder Framework


Motivation: cloud bot services vs bot as a web service models. We favor bot as a web service model for ease of use, control, extensibility etc.

MSFT bot builder framework - what is it?

Node.js and why we use the Node.js version vs the C# version

Core concepts

Intro to emulator

Node.js samples

How to run sample

Some basic bot samples


Chapter 5: Dive into Bot Builder


Motivation: we need to go over the different concepts inside the bot builder framework and how they help us build bots.

Dialogs

Messages

Connector

Channels

Conversation State

Recognizers

Actions

Proactive messages


Chapter 6: Putting it all together


Motivation: Build a sample bot that uses all of the concepts in the previous chapter

Back to ordering tickets bot - lots of coding

Chapter 7: Advanced bot builder topics

Motivation: Expand sample from #6 with some more advanced concepts like proactive messaging, better dialog management, custom recognizers, deeper action support, group chats, etc.

Lots of code here

Chapter 8: Deploying to Channels and sending custom data

Motivation: a bot is useless unless it is connected to a channel. Let's see how this is done. We introduce the bot framework web site.

Focus on Messenger and Slack

Show some custom functionality such as messenger square cards or slack updating messages


Chapter 9: Custom channels


Motivation: sometimes you would like to support a custom channel that is not supported out of the box by Microsoft. How do we do this?

Using direct line

Using vanilla bot builder interfaces


Chapter 10: Making the Bot smarter


Motivation: there is a lot of easy to use AI/ML services out there. What are they and how can they help our bot be smarter.




Vision

Emotion

Language Detection

Sentiment

Linguistic Analytics

QnA Maker




Why use these? Why not use these services? What is the goal of your bot?

Chapter 11: Communicating using Custom Graphics

Motivation: rich content and graphics can better capture information than plain text using dynamically generated images

Adaptive Cards - adaptivecards.io - one option, easier to deal with

Custom HTML generation - generate images from html using PhantomJS on the server


Chapter 12: Human Handoff


Motivation: customers want to keep in touch with your brand, what happens when a human is needed?


Chapter 13: Logging and Bot Analytics


How do we log data?

What are the interesting measures that we look for in a bot?

What systems are out there and what can they can us figure out about our customers and bots?


Chapter 14: Other bot implementation platforms


Brief overview of the technology field: Api.ai, wit.ai, IBM conversation, amazon, etc.

Goal of this chapter will be to demonstrate that these frameworks have very similar approaches in terms of NLU, conversation management, state, etc. Knowledge from this book will allow you to develop using the other systems.